Qwen2.5-VL
Qwen2.5-VL is Alibaba's open vision-language model family from early 2025. It is the base of several OCR models, including olmOCR 2 and Nanonets-OCR2.
A release of Qwen-VL. Other versions: Qwen3-VL, Qwen2-VL.
- Released
- January 28, 2025 (32B in March 2025)
- Model ID
- Qwen/Qwen2.5-VL-3B-Instruct, Qwen/Qwen2.5-VL-7B-Instruct, Qwen/Qwen2.5-VL-32B-Instruct, Qwen/Qwen2.5-VL-72B-Instruct
- Size
- 3B, 7B, 32B and 72B
- License
- Differs by size: 7B is Apache-2.0, 3B uses the Qwen Research license, 72B uses the Qwen license
- Also searched as
- qwen 2.5 vl, qwen2.5-vl, Qwen2.5-VL-7B-Instruct
What is new in Qwen2.5-VL
- Structured output from scanned invoices, forms and tables.
- Locates text and objects with bounding boxes or points and returns stable JSON.
- Higher document scores than Qwen2-VL at 7B: DocVQA 95.7 against 94.5, OCRBench 864 against 845.
Qwen2.5-VL benchmarks
| Benchmark | Score | Note |
|---|---|---|
| DocVQA (test) | 95.7 | 7B. Qwen2-VL-7B: 94.5 |
| OCRBench | 864 | 7B. Qwen2-VL-7B: 845, GPT-4o-mini: 785 |
| CC-OCR | 77.8 | 7B. Qwen2-VL-7B: 61.6 |
| ChartQA (test) | 87.3 | 7B. Qwen2-VL-7B: 83.0 |
Scores as published by the authors. We have not rerun them.
Use Qwen2.5-VL through an API
OCRRouter runs a different release of Qwen-VL as qwen/qwen-vl.
Qwen-VL versions
- Qwen-VL: all versions, price, speed and features on OCRRouter
- Qwen3-VL (September and October 2025)
- Qwen2.5-VL (January 28, 2025 (32B in March 2025)), this page
- Qwen2-VL (August 29, 2024)
Questions
Sources
Last updated .